Crystalline silicon surface passivation by high ... - AIP Publishing

JOURNAL OF APPLIED PHYSICS 107, 014504 共2010兲

Crystalline silicon surface passivation by high-frequency plasma-enhanced chemical-vapor-deposited nanocomposite silicon suboxides for solar cell applications Thomas Mueller,a兲 Stefan Schwertheim, and Wolfgang R. Fahrner University of Hagen, Chair of Electronic Devices, P.O.B. 940, Haldener Strasse 182, 58084 Hagen, Germany

共Received 5 May 2009; accepted 22 October 2009; published online 5 January 2010兲 A passivation scheme, featuring nanocomposite amorphous silicon suboxides 共a-SiOx : H兲 is investigated and analyzed in this work. The a-SiOx : H films are deposited by high-frequency plasma-enhanced chemical-vapor deposition via decomposition of silane 共SiH4兲, carbon dioxide 共CO2兲, and hydrogen 共H2兲 as source gases. The plasma deposition parameters of a-SiOx : H films are optimized in terms of effective lifetime, while the oxygen content and the resulting optical band gap EG of the a-SiOx : H films are controlled by varying the CO2 partial pressure ␹O = 关CO2兴 / 共关CO2兴 + 关SiH4兴兲. Postannealing at low temperatures of those films shows a beneficial effect in form of a drastic increase of the effective lifetime. This improvement of the passivation quality by low temperature annealing for the a-SiOx : H likely originates from defect reduction of the film close to the interface. Raman spectra reveal the existence of Si– 共OH兲x and Si–O–Si bonds after thermal annealing of the layers, leading to a higher effective lifetime, as it reduces the defect absorption of the suboxides. The surface passivation quality of a-SiOx : H within both n-type and p-type silicon has been studied as a function of injection level. Record high effective lifetime values of 4.7 ms on 1 ⍀ cm n-type float zone 共FZ兲 wafers and 14.2 ms on 130 ⍀ cm p-type FZ wafers prove the applicability for a surface passivation of silicon wafers applicable to any kind of silicon-based solar cells. The effective lifetime values achieved on a highly doped crystalline wafer 共1 ⍀ cm resistivity兲 appears to be the highest value ever reported. Samples prepared in this way feature a high quality passivation yielding effective lifetime values exceeding those of record SiO2 and SiNx passivation schemes. © 2010 American Institute of Physics. 关doi:10.1063/1.3264626兴 I. INTRODUCTION

In the field of crystalline silicon 共c-Si兲 solar cells, electronic surface passivation has been recognized as a crucial step to achieve high conversion efficiencies. High bulk and surface recombination rates are known to limit the open circuit voltage and to reduce the fill factor of photovoltaic devices.1,2 The suppression of surface recombination by applying any kind of surface passivation scheme is thereby one of the basic prerequisites to obtain high efficiency solar cells. Well surface-passivated solar cells tend to exhibit a higher open circuit voltage 共Voc兲. Therefore, in order to obtain high efficiency solar cells, it is essential to reduce the surface state density by passivating the c-Si surface properly. This becomes true for any silicon-based solar cell. Passivation schemes commonly used in photovoltaic applications are silicon dioxide 共SiO2兲,2 silicon nitride 共SiNx兲,3,4 but also intrinsic amorphous silicon 关a-Si: H共i兲兴,5 amorphous silicon carbide 共a-SiC: H兲,6 and stacks of those. The a-SiC: H is inferior compared to the a-Si: H共i兲, cf. Ref. 6. Thermally grown silicon oxide has shown excellent surface passivation properties 共cf. Ref. 2兲, resulting in a very low state density. However, the growth implies a high temperature application 共⬃1050 ° C兲, and it suffers from long term UV instability. Low-temperature processing sequences are based mainly on a兲

Electronic mail: [email protected].

0021-8979/2010/107共1兲/014504/11/$30.00

passivation with SiNx,3 a-Si: H,5 or a-SiC: H.6 The SiNx films are silicon rich and this fact brings along several drawbacks: the passivation quality depends strongly on the Si doping type and level; the films show a considerable absorption in the UV range of the solar spectrum, leading to a reduction of Jsc; the etch rate of those films is extremely low, hindering the local opening of the SiNx, which prohibit the applicability for heterojunction solar cells. The use of a-Si: H共i兲 has attracted the photovoltaic community due to the success of heterojunction with intrinsic thin-layer cells.7–9 The a-Si: H共i兲 films can be grown by plasmaenhanced chemical-vapor deposition 共PECVD兲 at low temperatures 共ⱕ200 ° C兲. However, Fujiwara and Kondo 共cf. Ref. 11兲 investigated a-Si: H / c-Si heterointerfaces, stating that they necessitate an immediate a-Si: H deposition and an abrupt and flat interface to the c-Si substrate. High temperature growth of a-Si: H共i兲, however, often leads to epitaxial layer formation on c-Si, which in turn reduces the solar cell performance severely. Thus, this epitaxial growth narrows the process window for the a-Si: H共i兲 layer growth and makes the solar cell optimization more difficult.10 Physical and structural properties of a-Si: H, however, strongly vary with the growth conditions. Also, due to the inherent strong blue light absorption, only ultrathin a-Si共i兲 : H films can be allowed to prevent losses. Originally focusing on the junction fabrication techniques of a-Si/ c-Si 共heterojunction兲 solar cells using a low-

107, 014504-1

© 2010 American Institute of Physics

014504-2

Mueller, Schwertheim, and Fahrner

temperature PECVD technique, passivation of the surface regions of the cell avoiding high-temperature cycling becomes an important issue. A passivation scheme, featuring high-transparent PECV-deposited hydrogenated nanocomposite amorphous silicon suboxides 共a-SiOx : H兲, grown at low temperatures, represents a material system suitable for this application and is quite an attractive alternative to standard a-Si: H incorporated in a-Si/ c-Si heterojunction solar cells. This passivation scheme using a-SiOx : H was already introduced in Ref. 12. This work was continued in Ref. 13 where we have proven the applicability of the PECVdeposited a-SiOx : H as a high quality surface passivation scheme for a c-Si based solar cell device. In this work we will review this work, and intensify our investigations on the a-SiOx : H films in terms of surface passivation quality. In particular, the optimization of the PECVD parameters will be associated to the compositional analysis of the resulting a-SiOx : H films, and the influence of stepwise postannealing of those films will be analyzed separately for both resistivity types. Furthermore, the passivation quality of the most effective a-SiOx : H films will be compared in various experiments, such as a comparison of the applied layer thickness of a-SiOx : H, a comparison to our most effective standard a-Si: H layers, as well as a comparison to different c-Si doping types and levels, superposed to record SiO2 and record SiNx passivation results published by Kerr and Cuevas 共cf. Refs. 2 and 3兲.

II. EXPERIMENTAL DETAILS

Intrinsic hydrogenated amorphous silicon 关a-Si: H共i兲兴 is generally grown by plasma decomposition of H2 and SiH4, whereas amorphous hydrogenated silicon suboxides 共a-SiOx : H兲 need an additional oxygen source. The decomposition of carbon dioxide 共CO2兲 as oxygen source, hydrogen 共H2兲, and silane 共SiH4兲 as precursor gases leads to the designated a-SiOx : H films. For this purpose a three chamber 共each chamber for either intrinsic, p-doped or n-doped layers to prevent contamination兲 very high-frequency PECVD setup is used. Therefore, the samples are placed into a 10 ⫻ 10 cm2 squared sample holder, which is suitable for up to 4 in. wafer substrates, and transferred via a load lock into one of the three chambers. The sample holder is attached to the upper 共electrically grounded兲 electrode, while the rf power is capacitively coupled to the lower electrode. A plasma power as high as 100 W can be adjusted. The rfPECVD system is driven by a rf generator, operating at frequencies ranging from 13.56 MHz up to very high frequencies at 110 MHz. Changing the excitation frequency does not necessarily lead to higher deposition rates but possible changes in the micromorph structure are likely. The distance between the parallel electrodes 共de兲 affects the emerging network of the deposited layers. Prior to deposition, the samples are heated up before being purged with argon gas. The samples are radiatively heated from above with actual substrate temperatures 共Tdep兲 being significantly lower than the adjusted heater temperature 共Theat兲. All samples for optical and electrical characterization are either deposited on high-

J. Appl. Phys. 107, 014504 共2010兲

quality silicon wafer material, oxidized wafer substrates or 7059 corning glass substrates, each individually precleaned. To determine optical constants and the corresponding optical band gap, EG, the films are deposited onto c-Si wafers covered with SiO2共1000 Å兲. Corning 7059 glass is used for deposition for ␮-Raman spectroscopy 共␮-RS兲. For lifetime measurements, the a-SiOx : H are directly deposited onto double-side polished floating zone 共FZ兲 wafers of different doping type and levels. The typical sample configuration to determine the passivation quality of the a-SiOx : H films is a symmetric a-SiOx : H / c-Si共FZ兲 / a-SiOx : H structure, equally treated on both sides. If needed, intrinsic amorphous silicon layers a-Si: H共i兲 are deposited for comparison at 13.56 MHz with identical process parameters, but without CO2. The plasma deposition parameters of a-SiOx : H films are optimized in terms of effective lifetime. The oxygen and hydrogen content, as well as the resulting optical band gap EG of the a-SiOx : H films are controlled by varying the CO2 partial pressure ␹O = 关CO2兴 / 共关CO2兴 + 关SiH4兴兲 and the H2 partial pressure ␹H = 关H2兴 / 共关H2兴 + 关SiH4兴兲. Therefore, all oxygen and hydrogen concentrations quoted in this work—unless stated otherwise—refer to the ratio ␹O and ␹H, respectively. To study the influence of annealing on the surface passivation quality and network bonding structure, the samples are annealed in a diffusion furnace under forming gas atmosphere 共10 at. % hydrogen diluted in nitrogen兲 at temperatures ranging from 100 to 500 ° C. The film compositions and also the changes in the microscopic structure of the amorphous network depending on plasma deposition frequencies and thermal annealing are studied by means of ␮-RS. For all ␮-Raman measurements presented in this section, the excitation is supplied by an Ar+ ion laser using a wavelength of 488 nm and a power of 25 mW. The spot size of the focused laser light is ⬃1 ␮m. The spectra are collected at room temperature by a Peltier cooled chargecoupled device detector. The resolution is limited to 0.7 cm−1. The spectra are recorded in the wave number range from 200 to 2500 cm−1. Secondary-ion-mass spectroscopy 共SIMS兲 gives information about the composition in the microscopic network, in particular the quantity of build-in oxygen and defective carbon atoms. Band gap analysis of the a-SiOx : H films depending upon used oxygen fraction during deposition is given at the end of this section using spectroscopic ellipsometry. The surface passivation quality of a-SiOx : H within both n-type and p-type silicon has been studied as a function of injection level. Therefore, investigations on the surface passivation quality of PECV-deposited a-SiOx : H films are carried out by measuring the effective carrier lifetime using quasisteady state 共QSS兲/transient photoconductance 共TPC兲 techniques. The QSS/TPC techniques provide a contactless measurement of the effective recombination lifetime of free carriers. The effective lifetime curves presented in this work are composed of multiple curves measured in QSSPC and TPC mode on the same axis. The simplified formula for a homogeneous, defect-free wafer with a spatially uniform carrier lifetime is given by

014504-3

J. Appl. Phys. 107, 014504 共2010兲


TABLE I. Varied PECVD parameters for the optimization of a-SiOx : H. Deposition parameter

Variation

Plasma frequency 共MHz兲 13.56–110 Plasma power 共W兲 2–20 Process pressure 共mTorr兲 200–1000 110–200 Deposition temperature Tdep共°C兲 0–50 Gas flow rate ␹O 共at. %兲 0–99.5 Gas flow rate ␹H 共at. %兲 Gas flow SiH4 关SCCM 共SCCM denotes cubic centimeter per minute at STP兲兴 1–40

1 1 1 = + , ␶eff ␶bulk ␶surf

共1兲

where ␶eff is the measured effective lifetime, ␶bulk is the bulk lifetime 关combined Auger, radiative, and Shockley–Read– Hall 共SRH兲 recombination兴, ␶surf is the characteristic surface recombination lifetime component determined by wafer thickness W, and surface recombination velocity 共SRV兲 Seff. From Eq. 共1兲 it can be concluded, that a better surface passivation leads to a higher ␶eff. For calculation of Seff it is assumed that both surfaces provide a sufficiently low recombination velocity and have the same values 共Seff = Sfront = Sback兲, as the sample structure is symmetric. Seff can be deduced by Seff =

冉

冊

1 W 1 − . 2 ␶eff ␶bulk

共2兲

The uncertainty of Seff depends on the value used for ␶bulk. Hereafter—unless stated otherwise—an upper limit of the SRV is calculated for the case that no SRH recombination is considered by setting ␶bulk → ⬁ in Eq. 共2兲. III. REVIEW OF PREVIOUS WORK A. Optimization of a-SiOx : H PECVD parameters

The PECVD parameters for amorphous silicon suboxide layers 共a-SiOx : H兲 are studied and optimized in terms of the corresponding surface recombination. Special attention has been paid to the effective lifetime achievable with PECV deposited a-SiOx : H. Evaluation of the SRV implies a double-sided c-Si wafer passivation. In this section, doubleside polished, 250-␮m-thick, 0.8– 1.2 ⍀ cm n-type and 0.475– 0.525 ⍀ cm p-type FZ wafers are used as silicon substrates, as substrates of this type, and resistivity can be used later on for solar cell device fabrication. As the bulk lifetime of each sample is comparable and negligible compared to the variations in the measured effective lifetime of different samples, those variations can be attributed to variations in the surface passivation. Thus, the process parameters that give the highest effective lifetime also give the lowest surface recombination velocity. All silicon substrates are RCA 共developed at Radio Corporation of America兲 cleaned prior to deposition, including a final etching in diluted HF to remove any oxide formed during RCA clean. The deposition parameters, as described in the previous section, will be optimized with respect to the following process variables, cf. Table I. The process parameter remaining is the deposition

FIG. 1. 共Color online兲 Effective lifetime as a function of annealing temperature depending on the plasma frequency for a-SiOx : H layer deposited at plasma frequencies of 13.56, 70, and 110 MHz. Two oxygen ratios are applied 共a兲 ␹O = 30% and 共b兲 ␹O = 50%. The chamber pressure and deposition temperature are set to 300 mTorr and 175 ° C, respectively. For comparison, the effective lifetime of intrinsic amorphous silicon without addition of oxygen 关a-Si: H共i兲兴 is added to the graph. The lines are a guide to the eye. A graph with higher solution for a plasma frequency of 70 MHz is given in Fig. 5.

time, which determines the a-SiOx : H layer thickness. A deposition time of 600 s is used for the optimization experiments. This corresponds to 100–250 nm thickness of the a-SiOx : H films, depending on the deposition rate. 1. Effect of plasma frequency

Initially, a-SiOx : H films are deposited at various plasma frequencies in the range of 13.56–110 MHz, and gas compositions containing oxygen fraction of ␹O = 0 – 50 at. %. In this experiment, the intrinsic a-SiOx : H layers are deposited at a heater temperature 共Theat兲 of 350 ° C, corresponding to a deposition temperature 共Tdep兲 of 175 ° C, and the chamber pressure is set to 300 mTorr. In Fig. 1, the effective lifetimes of samples deposited on p-type FZ wafers with plasma frequencies of 13.56, 70, and 110 MHz are shown. a-SiOx : H films of two different compositions are applied: 共i兲 ␹O = 30% and 共ii兲 ␹O = 50%. As seen from Fig. 1, the highest lifetimes are achieved with a plasma frequency of 70 MHz at annealing temperatures of 250 ° C. Annealing those samples at higher temperatures leads to an effusion of hydrogen, resulting in a detoriation of the surface passivation. The impact of thermal annealing is discussed separately and addressed in Sec. IV B 2. The effective lifetime of a-SiOx : H passivated c-Si deposited at 110 MHz shows an unexpected continuous increase of the effective lifetime at annealing temperatures Tann ⱖ 250 ° C 共up to 500 ° C兲, whereas in contrary the effective lifetime for samples of a-SiOx : H passivated c-Si deposited at lower plasma frequencies drops at that high temperatures around 500 ° C. The increase of the effective lifetime for samples deposited with 110 MHz might be due to a microcrystalline character of the a-SiOx : H 共discussed in Sec. IV B兲, preventing a hydrogen effusion during annealing, as well as due to further diffusion of hydrogen to the c-Si interface.

014504-4


J. Appl. Phys. 107, 014504 共2010兲

FIG. 2. 共Color online兲 Effective lifetime 共␶eff兲 as a function of oxygen content 共␹O兲 in the feed stock during the a-SiOx : H decomposition. For each oxygen content, the effect of postannealing is shown for temperatures of 250 and 500 ° C. The a-SiOx : H films are deposited with Tdep = 170 ° C and nonoptimized ␹H and SiH4 gas flow rates.

FIG. 3. 共Color online兲 Effective lifetime 共␶eff兲 as a function of the heater temperature 共Theat兲 before and after postanealing at 250 ° C. A maximum ␶eff is found at a Theat = 320 ° C, corresponding to a Tdep = 155 ° C. The a-SiOx : H films are deposited with ␹O = 20 at. %, but nonoptimized ␹H and SiH4 gas flow rates.

2. Effect of gas flow rates

4. Effect of preheating time

In Fig. 2 the effective lifetime of a-SiOx : H passivated c-Si samples is displayed as a function of oxygen content 共␹O兲 in the feedstock during deposition. For this experiment, the deposition temperature 共Tdep兲 has been set to 170 ° C and the plasma frequency to 70 MHz. As seen from Fig. 2, with increasing ␹O from 0 to 20 at. %, the ␶eff increases due to a passivating effect of the oxygen build into the resulting amorphous network during deposition. The highest effective lifetimes of ␶eff ⱖ 2 ms 共corresponding to a low Seff = 6.6 cm/ s兲 is obtained for an oxygen content of ␹O = 20 at. %. With an oxygen content higher than ␹O = 20 at. % in the gas phase, the lifetime decreases. One possible explanation could be that even though the content of carbon contained in the CO2 is assumed to be below 1%, the carbon fraction increases with increasing oxygen content, resulting in a higher defect density. This effect will be investigated further in the following section 共cf. Sec. IV B兲, which is dealing with SIMS measurements.

Initially, the a-SiOx : H films are optimized for a preheating time of theat = 15 min. To study the dependence on temperature changes in the chamber during preheating, the PECVD process parameters are kept constant but the preheating time, theat, is varied from theat = 15– 120 min. Figure 4 illustrates the dependency of ␶eff on theat of the c-Si sample in the PECVD setup for theat = 15– 120 min. It is evident that the preheating time becomes a critical factor, as the deposition temperature 共Tdep兲 is not constant, and can be expressed as a function of theat. A theat of 15 min corresponds to a deposition temperature Tdep of less than 155 ° C, hence with increasing theat time, Tdep increases. This increase of Tdep results in a change of the a-SiOx : H decomposition. As seen already in Fig. 3, Tdep needs to be exactly 155 ° C for optimal lifetime results. The choice of a short theat and corresponding Tdep, rather than a longer theat and lower Tdep, is purely based on process time within our laboratory.

3. Effect of deposition temperature

A variation in the deposition temperature Tdep 共or referred to as heater temperature Theat兲, will be discussed in the following. In Sec. III A 2, an optimum oxygen content of ␹O = 20 at. % has been found in respect of the obtained effective lifetime. Therefore, for this experiment investigating the deposition temperature dependence, the oxygen content is kept constant at ␹O = 20 at. %. The plasma frequency is set to 70 MHz. Figure 3 shows the effective lifetime of a-SiOx : H passivated c-Si samples as a function of the heater temperature 共Theat兲. The highest effective lifetime is obtained for samples deposited at Theat = 320 ° C 共corresponding to a deposition temperature Tdep = 155 ° C兲, as ␶eff ⱖ 2 ms, corresponding to Seff ⱕ 6.6 cm/ s, after subsequent annealing at 250 ° C. Thermal annealing at higher temperatures 共60 min at 500 ° C兲 leads to a drastic decrease in the effective lifetime for all samples the samples due to the effusion of hydrogen 共the impact of postannealing of the a-SiOx : H on the effective lifetime is discussed in detail in Sec. IV B 2兲.

FIG. 4. 共Color online兲 Impact of the preheat time 共theat兲 of the c-Si sample in the PECVD setup prior deposition on the effective lifetime of a-SiOx : H passivated samples. The effective lifetime decreases with increasing theat due to a change in the resulting micromorph a-SiOx : H structure. Apart from the variation in theat in this experiment, the process parameters for deposition of a-SiOx : H are optimized.

014504-5

J. Appl. Phys. 107, 014504 共2010兲


TABLE II. Optimized deposition conditions for the developed a-SiOx : H films. Process parameter Gas flow rate ␹O Deposition rate 共Å/s兲 Plasma frequency 共MHz兲 Tdep共°C兲 Electrode distance 共mm兲 Postannealing

FIG. 5. 共Color online兲 Influence of stepwise annealing Tann step on the electronic passivation quality, expressed by the effective lifetime ␶eff and the SRV Seff of thin intrinsic 共optimized兲 a-SiOx : H layers deposited on mirror polished 1 ⍀ cm n-type FZ-Si and 0.5 ⍀ cm p-type FZ-Si surfaces. The lines are guides for the eye.

IV. RESULTS AND DISCUSSION A. Impact of postannealing on passivation properties of a-SiOx : H / c-Si heterostructures

To investigate the influence of postannealing of the a-SiOx : H films for both c-Si resistivity types separately, the a-SiOx : H films are deposited on both 1 ⍀ cm n-type and 0.5 ⍀ cm p-type mirror polished FZ-Si substrates. After that, the samples are stepwise annealed in a diffusion furnace in a range of 100– 500 ° C under forming gas atmosphere 共10% H2 in nitrogen兲 for 3 h. For this sample series the plasma frequency and oxygen fraction are set to 70 MHz 共corresponding to the highest lifetime values of samples presented in Fig. 1兲 and ␹O = 20%, respectively. Figure 5 shows the changes of the electronic passivation quality due to stepwise postannealing of the a-SiOx : H films deposited on both n-type and p-type c-Si. For the n-type Si-FZ it is found that subsequent annealing of the samples at 250 ° C for 3 h in forming gas atmosphere increases drastically the effective lifetime to its maximum. A high effective lifetime 共␶eff = 4.7 ms, leading to Seff ⱕ 2 cm/ s兲 has been achieved for samples deposited with an oxygen content of 20 at. % after subsequent annealing at 250 ° C. This might be 共i兲 due to hydrogen saturation of Si dangling bonds at the a-SiOx : H surface, or 共ii兲 due to the generation of Si-共OH兲x and Si–O–Si bonds in the microscopic structure of the amorphous network 共cf. Raman analysis in Sec. IV B兲. Thermal annealing at higher temperature 共up to 500 ° C兲 leads to a steady decrease in the effective lifetime of the samples due to 共i兲 an effusion of hydrogen or 共ii兲 an increase in carbon with increasing oxygen content from the carbon dioxide, resulting in a higher defect density. Also, for a-Si: H共i兲 material annealed at higher temperatures, a correspondence between the hydrogen effusion rate and defect generation in the film has been demonstrated in the past by comparing thermal desorption spectroscopy 共TDS兲, infrared absorption, and electron spin resonance measure-

a-SiOx : H deposition conditions 1:5 3.5 70 155 19 250 ° C for n-type c-Si 300 ° C for p-type c-Si

ments 共cf. Ref. 14兲. The same mechanism can be attributed to a-SiOx : H, where hydrogen likely is transferred at temperatures above 350 ° C from a Si–H to a H2 state, creating defects in the material 共cf. Sec. IV B兲. As stated in Ref. 15, the quality of a passivation scheme applied to a silicon material of different conduction type differs. This is consistent for the a-SiOx : H passivated p-type Si-FZ substrates: resulting in a lower effective lifetime of ␶eff ⱖ 2 ms and a shift of the maximum of ␶eff toward 300 ° C 共compared to a-SiOx : H passivated n-type material with a maximum of ␶eff around 250 ° C annealing temperature兲. The lower lifetime of p-type passivated material 共␶eff ⱖ 2 ms兲 compared to n-type passivated material 共␶eff ⱖ 4 ms兲 might be attributed mainly to the defect states in our nonstoichiometric a-SiOx : H films. It can be assumed that some of these defect states are within the band gap of the c-Si at the interface, and therefore directly induce the recombination at the interface to the c-Si material of different conduction type. In conclusion, the PECVD plasma parameters are optimized in terms of measured effective lifetime. Table II displays these optimized deposition parameters developed in this section. Hereafter, these parameters are used to prepare high quality a-SiOx : H films, unless stated otherwise.

FIG. 6. 共Color online兲 Raman spectra of a-SiOx : H films directly after deposition using plasma frequencies of 13.56, 70, and 110 MHz. Full range between 200 and 1600 cm−1. The inset exhibits a close-up range between 400 and 500 cm−1. For comparison the spectra of intrinsic a-Si: H共i兲 are superposed.

014504-6


J. Appl. Phys. 107, 014504 共2010兲

FIG. 7. 共Color online兲 Raman spectra of a-SiOx : H films deposited at 70 MHz: the spectra of a sample directly after deposition and after thermal annealing at 250 ° C are superposed. A generation of Si–O–Si bonds as well as Si共OH兲x bonds is observed.

FIG. 8. 共Color online兲 Raman spectra after thermal annealing at 250 ° C deposited using plasma frequencies of 13.56, 70, and 110 MHz. For comparison the spectra of intrinsic a-Si: H共i兲 after thermal annealing are superposed.

B. Compositional analysis of a-SiOx : H

However, the generation of Si–O–Si and Si– 共OH兲x bonds due to thermal annealing does not depend on the applied plasma frequency, as can be seen in the following. Figure 8 show the Raman spectra for a-SiOx : H films deposited at 13.56, 70, and 110 MHz after thermal annealing at 250 ° C superposed to an a-SiOx : H film deposited at 70 MHz directly after deposition. The same conclusions for samples deposited with plasma frequencies of 13.56 and 110 MHz can be drawn as observed for samples deposited at 70 MHz. For samples deposited at 110 MHz, the generation of Si– O–Si bonds and Si– 共OH兲x bonds appears to be less pronounced compared to the samples deposited with 13.56 and 70 MHz because the intensity is typically measured in arbitrary units. As shown in Sec. IV A, subsequent annealing of the samples drastically increases the effective lifetime. This might be 共i兲 due to hydrogen saturation of Si dangling bonds at the a-SiOx : H surface, or 共ii兲 due to compositional changes in the microscopic structure of the amorphous network. In case of standard a-Si: H共i兲共␹O = 0 at. %兲, the hydrogen content is roughly 12–14 at. % as reported by Refs. 19 and 20. It is also known that the amorphous silicon suboxides contain a

1. Changes in the microscopic structure upon plasma frequency

Changes in the microscopic structure upon plasma frequency for deposition are investigated using Raman spectroscopy. Figure 6 displays the corresponding Raman spectra of films deposited with plasma frequencies at 13.56, 70, and 110 MHz. In addition, the spectra of the reference a-Si: H共i兲 layer are superposed as a reference spectrum. As seen from Fig. 6, the character of films deposited at 13.56 MHz can be attributed to amorphous, whereas the character of films deposited 70 MHz is close to the transition to microcrystalline 共nanocomposite兲 around 500 cm−1. The spectra of the films deposited at 110 MHz exhibit a pronounced peak at 500 cm−1 corresponding to a microcrystalline character.

2. Changes in the microscopic structure upon thermal annealing

Changes in the microscopic structure upon thermal annealing are investigated using Raman spectroscopy. Figure 7 exhibit the Raman spectra for an a-SiOx : H film deposited at 70 MHz, directly after deposition and after subsequent postannealing at 250 ° C. From Fig. 7 can be concluded that after postannealing of the film, a generation of Si-共OH兲x bonds appear with a peak formation around 920 and 1079 cm−1, which agrees well with the results shown by Refs. 16 and 17. Furthermore, Si–O–Si bonds occur around 820 cm−1, whereas the reference sample a-Si: H共i兲 does not show any compositional changes due to thermal annealing in the microscopic structure. For samples annealed at higher temperatures 共Tann = 500 ° C for 1 h兲 significantly lower Si– 共OH兲x or Si–O–Si peaks are detected 共not shown in the figures兲. Effusion of hydrogen at temperatures around 400– 500 ° C might be the reason, as reported by Ref. 18. Therefore, one can conclude that plasma frequencies influence only the amorphous/microcrystalline character during deposition, and show no effect on the generation of bonds like Si– 共OH兲x or Si–O–Si after thermal annealing.

FIG. 9. 共Color online兲 SIMS spectra of samples deposited with different ␹O rates exhibiting the oxygen content in the a-SiOx : H film. Taken from Ref. 24.

014504-7

J. Appl. Phys. 107, 014504 共2010兲


FIG. 10. 共Color online兲 SIMS spectra of samples deposited with different ␹O rates exhibiting the carbon content in the a-SiOx : H film. The concentration of carbon appears to be less than 1 vol %. Taken from Ref. 24.

significantly higher fraction of hydrogen. Raman spectra around the 2000 cm−1 band region show typical peaks for Si–H bondings for all prepared a-SiOx : H samples 共not depicted in the figures兲. Therefore, it can be assumed that the fraction of hydrogen in the films is sufficiently high to cover all dangling bonds and that no additional hydrogen is built up in the amorphous network from annealing in forming gas. The concentration of carbon and other impurities is below 1 at. % for all samples investigated in this section. Nevertheless, lattice distortions, such as SiO2 or Si clusters, silicon dangling bonds, molecular hydrogen inclusions, or microvoids with hydrogen terminated internal surfaces are likely to exist in the a-SiOx : H samples. 3. SIMS analysis of a-SiOx : H films

SIMS has been carried out to analyze the composition of the a-SiOx : H films. Explicitly, measurements of films with an oxygen content of ␹O = 0, 20, and 50 at. % are compared and depicted in Figs. 9 and 10. The choice of the ␹O ratio is based on the lifetime results previously achieved, where the sample with ␹O = 20 at. % showed the highest effective lifetime, the sample with ␹O = 0 at. % is used as reference and the sample with ␹O = 50 at.% represents the highest investigated oxygen content. It is found that oxygen is built into the amorphous network; with increasing oxygen content in the feed stock during deposition, SIMS reveals an increasing oxygen content in the resulting network. A higher oxygen content leads also to an increased deposition rate 共from 3.0 up to 3.8 Å/s兲, indicated by the interface peaks. The pileups visible at the interface air/ a-SiOx : H and at the interface a-SiOx : H / c-Si can be explained as segregation effects. Furthermore, the carbon fraction in the a-SiOx : H films increases with higher ␹O content during deposition. This result explains the decrease in the effective lifetime at ␹O = 50 at. % 共shown in Fig. 2兲; indicating a higher defect density.

FIG. 11. 共Color online兲 Absorption coefficient ␣ of a-SiOx : H deduced from SE data fit compared to published absorption coefficients of crystalline silicon 共cf. Jellison 3–1-91兲 and a-Si:H 共cf. Palik HOC I, 577–580兲. The absorption coefficient at the blue light region around 3.0 eV is by one order of magnitude lower than the one of standard a-Si: H. Taken from Ref. 13.

a-SiOx : H films are deposited on a c-Si/ SiO2 substrate; the 1000-Å-thick SiO2 layer is then employed as an optical separation layer to enhance the contrast. For the SE analysis, a simple three layer model c-Si/ SiO2 / a-SiOx : H is used. The deposited thickness is determined by profilometer measurements. The absorption coefficient ␣ is then calculated by means of

␣=

4␲k . ␭

共3兲

Figure 11 共taken from Ref. 13兲 shows ␣ as a function of the photon energy for the a-SiOx : H films compared to that of standard a-Si: H共i兲 and crystalline silicon. The absorption in the a-SiOx : H film at the blue light region around 3.0 eV is significantly lower than that of a-Si: H共i兲. Consistently, the fraction of blue light transmitted to the wafer which is available for solar conversion increases drastically 共though depending on the applied thickness兲. As the passivation quality depends strongly on the thickness of an passivating layer,21,22 it is evident that with decreasing absorption in the passivation layer, the thickness could be increased to gain a better passivation quality. The surface passivation depending on applied layer thickness of a-SiOx : H will be investigated in Sec. IV C 4 in detail. The optical band gap EG of the a-SiOx : H films strongly depends on ␹O and linearly rises with increasing oxygen content from 1.7 eV 共␹O = 0 at. %兲 up to 2.4 eV 共␹O = 50 at. %兲, cf. Fig. 12. Thermal annealing at 250 ° C does influence the resulting band gap. As has been discussed earlier the annealing at higher temperatures 共500 ° C兲 leads to a decrease in EG 共not depicted兲, which can be as well attributed to the effusion of hydrogen. The optical band gaps of samples deposited with 110 MHz are constant even after annealing at higher temperatures 共500 ° C兲, due to the microcrystalline structure.

4. Optical confinement of a-SiOx : H layers

From spectroscopic ellipsometry 共SE兲 measurements of the a-SiOx : H films, the real part of the refractive index n and the extinction coefficient k can be deduced. For this purpose,

C. Surface passivation quality of a-SiOx : H

In the previous section 共cf. Sec. III A兲, the optimization of the PECVD parameters of the a-SiOx : H layers is de-

014504-8


FIG. 12. 共Color online兲 Optical band gap of a-SiOx : H films deposited at 70 MHz as a function of ␹O; after deposition and after postannealing at 250 ° C. The sample configuration used for the band gap determination is displayed in the inset.

scribed. This section refers to the surface passivation quality of a-SiOx : H compared to passivation quality of various passivation schemes published elsewhere, among those the standard SiO2, SiNx, and a-Si: H共i兲. Furthermore, this section contains a study of a-SiOx : H passivating c-Si of various doping levels and types. The homogeneous distribution of the measured ␶eff will be investigated, as well as the dependency of the applied a-SiOx : H layer thickness in respect of the application to heterojunction solar cell devices in stacks with a-Si: H films are compared. Comparing different passivation schemes such as SiO2 and SiNx published elsewhere implies identically measurement conditions as well as identically analysis of the measured data. Passivation quality depends on applied film thickness of the passivating film and the doping type and level of the material used. Deciding on the specified mean carrier density 共MCD兲 specifying a value for ␶eff is therefore discretionary. Ideally, a MCD is chosen that corresponds to the Vmp in the used material. However, 1 ⫻ 1015 cm−3 is very appealing because it is well above the noise, above trapping, and above dual resonance model effects. Some authors tend to use the maximum occurring ␶eff, some prefer the 1 ⫻ 1015 cm−3. To ensure comparability, a MCD of 1 ⫻ 1015 cm−3 is used to specify the value of ␶eff hereafter. As the sample structure is symmetrical, all Seff values presented hereafter are calculated under the assumption that both surfaces provide a sufficiently low recombination velocity and have the same values 共Seff = Sfront = Sback兲. One should note that the uncertainty of Seff depends strongly on the value used for ␶bulk. Therefore, an upper limit of the SRV is calculated for the case that 共i兲 no SRH recombination is considered 共by setting ␶bulk → ⬁ in Eq. 共2兲, leading to an approxi−1 兲, with W defining the wafer mate value of Seff = W ⫻ 共2 ⫻ ␶eff thickness兲 and 共ii兲 for ␶bulk = 35 ms after,23 which is well above the highest ever measured effective lifetime. 1. Surface passivation quality of a-SiOx : H compared to standard a-Si: H„i…

As for heterojunction solar cells, a-Si: H共i兲 is the standard material applicable as buffer layer between c-Si and

J. Appl. Phys. 107, 014504 共2010兲

FIG. 13. 共Color online兲 Measured effective lifetime for a-SiOx : H passivated n-type FZ wafers and standard intrinsic a-Si: H共i兲 passivated n-type FZ wafers as a function of injection level. Using optimized deposition parameters, the value for ␶eff increased up to 4.7 ms 共at 1 ⫻ 1015 cm−3兲 for postannealed a-SiOx : H deposited on 1 ⍀ cm n-type FZ. In comparison, for the best a-Si: H共i兲 layer, a value for ␶eff of 2.1 ms has been obtained.

doped a-Si: H, as well as surface passivation of the c-Si material. Therefore, the surface passivation quality of a-SiOx : H is compared to the passivation quality of standard a-Si: H共i兲 in this experiment. The passivation quality is characterized by the effective lifetime on 1 ⍀ cm n-type FZ material. Both a-SiOx : H and a-Si: H共i兲 are deposited on front and backside with the equal, optimized process conditions. Using optimized deposition parameters, an excellent effective lifetime of 4.1 ms directly after deposition of a-SiOx : H on a 1 ⍀ cm n-type FZ wafer has been achieved, as illustrated in Fig. 13. After postannealing at 250 ° C for a duration of 3 h in hydrogenated atmosphere the ␶eff increased to 4.7 ms. This is by far the highest ever measured effective lifetime value for a PECV-deposited amorphous silicon passivated, high-doped 共1 ⍀ cm兲 n-type wafer. In contrary, effective lifetime values for our best a-Si: H共i兲 films, are 1.7 ms after deposition and 2.1 ms after postannealing at 250 ° C. The results for the a-Si: H共i兲 films passivating the 1 ⍀ cm n-type wafer are consistent with other publications. Table III summarizes the obtained values for ␶eff and Seff. 2. Surface passivation quality of a-SiOx : H compared to standard SiO2 and SiNx passivation schemes

Up to now, record values of effective lifetimes are established for SiO2 and SiNx by Kerr and Cuevas published in Refs. 2 and 3 the highest effective lifetimes 共␶eff兲 which incorporates all the bulk and surface recombination processes, previously reported for crystalline silicon appears to be ␶eff = 29 and 32 ms, corresponding to the lowest Seff with 0.46 and 0.625 cm/s for a 90 ⍀ cm FZ n-type material and a 150 ⍀ cm FZ p-type material, respectively, passivated with annealed SiO2. Note, that the values given are obtained using the effective lifetime at its maximum value, neglecting the choice of a discretionary injection level. However, the SRV for higher doped material, i.e., 1.5 and 0.6 ⍀ cm material, increased to 2.4 and 26.8 cm/s, respectively, for wafers passivated with SiO2, as shown in Fig. 14.

014504-9

J. Appl. Phys. 107, 014504 共2010兲


TABLE III. Results for a-SiOx : H and a-Si: H共i兲 passivated c-Si wafers: maximum measured effective lifetimes, ␶eff, at 1 ⫻ 1015 cm−3, and the corresponding SRV for an upper limit 共␶bulk → ⬁兲 and for ␶bulk = 35 ms.

Dopant type

Doping level 共⍀ cm兲

Doping concentration 共cm−3兲

Passivation method

␶eff 共1 ⫻ 1015 cm−3兲共ms兲

SRV 共␶bulk → ⬁兲共cm/s兲

SRV 共␶bulk = 35 ms兲共cm/s兲

n-type n-type n-type n-type

1 1 1 1

5 ⫻ 1015 5 ⫻ 1015 5 ⫻ 1015 5 ⫻ 1015

a-Si: H共i兲 a-Si: H共i兲, annealed a-SiOx : H a-SiOx : H, annealed

1.7 2.1 4.1 4.7

7.35 5.95 3.05 2.65

6.99 5.59 2.69 2.30

Comparing those results to our passivated 1 ⍀ cm n-type FZ with a-SiOx : H, a SRV of outstanding 2.6 cm/s using a PECV grown a-SiOx : H passivation scheme is reached 共assuming that ␶bulk → ⬁兲, see Fig. 14. In Fig. 14, this record SiO2 passivation of Kerr and Cuevas2,3 is superposed to the QSSPC results of a-SiOx : H on 1 ⍀ cm FZ material. An excellent ␶eff of ⱖ4 ms on 1 ⍀ cm material is repeatedly reached and confirmed by different QSSPC setups. This result appears to be the highest measured effective lifetime for 1 ⍀cm passivated c-Si material. 3. Surface passivation quality for bulk material of different doping type and level

The most research publications dealing with surface passivation schemes use low-doped substrates, stating that those wafers contain less defect states, although a subsequent solar cell device is fabricated on high-doped wafer around 1 ⍀ cm. The results achieved in this work with a-SiOx : H as a high quality surface passivation scheme are compared to other passivation schemes, by applying a-SiOx : H films to lower doped wafer substrates. The effective lifetime measured by QSSPC and TPC depends strongly on the doping level and type used. A selection of a 130 ⍀ cm p-type FZ wafer, as well as 60 and 1 ⍀ cm n-type FZ wafers, which have been passivated using a-SiOx : H with subsequent annealing, are presented. The choice of FZ material with 60 and 130 ⍀ cm resistivity is purely based on wafer availability within our labo-

FIG. 14. 共Color online兲 Measured effective lifetime ␶eff for a-SiOx : H passivated n-type FZ wafers as a function of the excess carrier density in the range of 1012 – 1017 cm−3. Auger recombination dominates at the higher carrier injection levels. Previously reported passivation schemes of SiO2 and SiNx films by Kerr and Cuevas published in Refs. 2 and 3 are superposed.

ratory, as those wafers are sourced from commercial stock. The results of the measured ␶eff and corresponding Seff are presented in Fig. 15. A maximum effective lifetime of 14.2 ms has been measured for lightly doped p-type material at an injection level of 1 ⫻ 1015 cm−3. This is by far the highest effective lifetime ever measured for a PECV-deposited amorphous silicon passivated wafer. In Table IV the results of our passivation scheme are listed in detail. To compare those excellent results further to previously published results by Kerr and Cuevas 共cf. Refs. 2 and 3兲, Figs. 16 and 17 demonstrate the surface passivation quality of a-SiOx : H films compared once more to SiO2 and SiNx passivation schemes depending on different doping types and levels. 4. Surface passivation quality depending on applied a-SiOx : H film thickness

Here, the surface passivation quality of a-SiOx : H films depending on the applied film thickness is investigated. For better comparability, a 1 ⍀ cm n-type FZ wafer is used for all depositions. The thicknesses of the deposited films vary from 250 down to 2 nm, which is the lowest thickness ensuring a homogeneous, shunt-free layer distribution. The results of the effective lifetime ␶eff and corresponding Seff values for an a-SiOx : H passivated 1 ⍀ cm n-type wafer are shown in Fig. 18. A film thickness as low as 10 nm 共deposition time 35 s at a rate of 3 Å/s兲 provides an almost as good surface passivation quality as a 250-nm-thick layer. Further reduction of the thickness down to 3 nm results in a slight

FIG. 15. 共Color online兲 Measured effective lifetime for a-SiOx : H passivated n-type and p-type FZ wafers of various doping concentrations as a function of injection level.

014504-10

J. Appl. Phys. 107, 014504 共2010兲


TABLE IV. Results for a-SiOx : H passivated silicon wafers, depending on dopant type and doping level: maximum measured effective lifetimes, ␶eff at 1 ⫻ 1015 cm−3, and SRV for an upper limit 共␶bulk → ⬁兲 and for ␶bulk = 35 ms.

Dopant type

Doping level 共⍀ cm兲

Doping concentration 共cm−3兲

Passivation method

␶eff 共1 ⫻ 1015 cm−3兲共ms兲

SRV 共␶bulk → ⬁兲共cm/s兲

SRV 共␶bulk = 35 ms兲共cm/s兲

n-type n-type p-type

1 60 130

5 ⫻ 1015 7 ⫻ 1013 1 ⫻ 1014

a-SiOx : H a-SiOx : H a-SiOx : H

4.7 7.2 14.2

2.66 1.73 0.88

2.3 1.37 0.52

decrease in the effective lifetime to 3.1 ms. For a film thickness lower than 2 nm the a-SiOx : H films are not deposited homogeneously onto the wafer surface in the PECVD setup. Therefore, the effective lifetime decreases drastically and those results are neglected in Fig. 18. V. CONCLUSIONS

A passivation scheme, featuring PECV-deposited hydrogenated nanocomposite amorphous silicon suboxide films 共a-SiOx : H兲, has been investigated. The plasma deposition parameters of a-SiOx : H films are optimized in terms of effective lifetime, whereas the oxygen content and the resulting optical band gap EG of the a-SiOx : H films are controlled by varying the CO2 and H2 partial pressure. An optimum substrate deposition temperature and plasma frequency of Tdep = 155 ° C and 70 MHz, respectively, have been revealed. Also, an optimized gas flow ratio of ␹O = 1 : 5 has been employed. Additional hydrogen reduces the oxygen content to 10 at. % in the gas flow during deposition. The impact of postannealing at low temperatures around 250– 300 ° C of the a-SiOx : H films has shown a beneficial effect in form of a significant increase in the effective lifetime. The improvement of the passivation quality driven by low temperature annealing of the a-SiOx : H likely originates from defect reduction of the film close to the interface. Raman spectra reveal the existence of Si– 共OH兲x and Si–O–Si bonds after thermal annealing of the layers, leading to a higher effective lifetime, as it reduces the defect absorption

FIG. 16. 共Color online兲 Measured effective lifetime for a-SiOx : H passivated n-type FZ wafers of various doping concentrations as a function of injection level. Previously reported results for passivation schemes of SiO2 and SiNx films published in Refs. 2 and 3 are superposed.

of the suboxides. Thermal annealing at higher temperature 共above 500 ° C兲 results in a decrease in lifetime and to hydrogen effusion. A high-transparent window layer up to 2.4 eV, depending on the oxygen fraction in the precursor gas, is obtained by adding oxygen to the a-SiOx : H layers. The a-SiOx : H films exhibit a very low blue light absorption compared to a-Si: H共i兲, enabling a thicker passivation layer than standard a-Si共i兲 : H and therefore an increasing passivation quality. The surface passivation quality of a-SiOx : H within both n-type and p-type silicon has been studied as a function of injection level. Record high effective lifetime values of 4.7 ms on 1 ⍀ cm n-type FZ wafers and 14.2 ms on 130 ⍀ cm p-type FZ wafers prove the applicability for a surface passivation of silicon wafers to any kind of silicon-based solar cells. The values achieved in this work appear to be the highest ever reported values of ␶eff on a high doped crystalline silicon wafer of 1 ⍀ cm resistivity. Additionally, the a-SiOx : H films yield a surface passivation quality exceeding earlier published record passivation schemes such as SiNx and SiO2. Therefore, the use of a-SiOx : H may be a promising alternative for any passivation scheme existing so far. Advantages of the a-SiOx : H passivation scheme are that the fabricated a-SiOx : H layers are grown by simple PECVD at low temperatures, they withstand hydrofluoric acid 共tested with a 5% diluted HF dip for 1 min兲 and high temperatures up to 500 ° C 共tested for 1 h under nitrogen atmosphere兲. It has to be mentioned that the process parameters have to be

FIG. 17. 共Color online兲 Measured effective lifetime for a-SiOx : H passivated p-type FZ wafers of various doping concentrations as a function of injection level. Previously reported results for passivation schemes of SiO2 and SiNx films published in Refs. 2 and 3 are superposed.

014504-11


FIG. 18. 共Color online兲 Values for ␶eff and Seff as a function of applied a-SiOx : H film thickness, da-SiOx:H, extracted from QSSPC and TPC measurements at an excess carrier density of 1 ⫻ 1015 cm−3. For comparison, films directly after deposition and after annealing are superimposed. The lines are a guide to the eye.

carefully adjusted to obtain such high quality a-SiOx : H films. Main key factors to obtain such high quality films comprehend: 共i兲 the chamber contamination, which depend on anteceded processes with various precursor gases, 共ii兲 dummy process to cover the reactor walls with a-SiOx : H, 共iii兲 the deposition temperature and the preheating time in the chamber, and 共iv兲 the electrode distance. The maximum surface area for c-Si passivation with a-SiOx : H in this work has been limited to 10⫻ 10 cm due to the PECVD chamber size. For industrial processes the PECVD on of a-SiOx : H on large area is still a challenge, especially regarding to the optimized plasma frequency of 70 MHz. ACKNOWLEDGMENTS

We wish to acknowledge the support of Boguslaw Wdowiak 共University of Hagen兲 for various measurements, Katrina Meusinger 共University of Hagen兲 for sample preparation, and Uwe Zastrow 共FZ Juelich兲 for taking SIMS measurements. M. A. Green, Silicon Solar Cells-Advanced Principles and Practice 共Cen-

1

J. Appl. Phys. 107, 014504 共2010兲 tre for Photovoltaic Devices and Systems, University of South Wales, Sydney, 1995兲. 2 M. J. Kerr and A. Cuevas, Semicond. Sci. Technol. 17, 35 共2002兲. 3 M. J. Kerr and A. Cuevas, Semicond. Sci. Technol. 17, 166 共2002兲. 4 J. Schmidt, Ph.D. thesis, Universitaet Hannover, 1998. 5 M. Taguchi, K. Kawamoto, S. Tsuge, T. Baba, H. Sakata, M. Morizane, K. Uchihashi, N. Nakamura, S. Kiyama, and O. Oota, Prog. Photovoltaics 8, 503 共2000兲. 6 I. Martin, M. Vetter, A. Qorpella, J. Puigdollers, A. Cuevas, and R. Alcubilla, Appl. Phys. Lett. 79, 2199 共2001兲. 7 M. Taguchi, A. Terakawa, E. Maruyama, and M. Tanaka, Prog. Photovoltaics 13, 481 共2005兲. 8 M. Tanaka, M. Taguchi, T. Matsuyama, T. Sawada, S. Tsuda, S. Nakano, H. Hanafusa, and Y. Kuwano, Jpn. J. Appl. Phys., Part 1 31, 3518 共1992兲. 9 E. Maruyama, A. Terakawa, M. Taguchi, Y. Yoshimine, D. Ide, T. Baba, M. Shima, H. Sakata, and M. Tanaka, Proceedings of the Fourth IEEE World Conference on Photovoltaic Energy Conversion 共WCPEC-4兲, Waikoloa, HI, 2006, p. 1455. 10 T. Wang, E. Iwaniczko, M. Page, D. Levi, Y. Yan, V. Yelundur, H. Branz, A. Rohatgi, and Q. Wang, Proceedings of the 31st IEEE Photovoltaic Specialists Conference 共IEEE-PVSC-31兲, Orlando, FL, 2005, p. 955. 11 H. Fujiwara and M. Kondo, J. Appl. Phys. 101, 054516 共2007兲. 12 T. Mueller, W. Duengen, R. Job, M. Scherff, and W. R. Fahrner, Proceedings of the Materials Research Society Symposium A Proceedings, 2006 MRS Fall Meeting, Boston, MA, 2007, p. 0989–A05–02. 13 T. Mueller, S. Schwertheim, M. Scherff, and W. R. Fahrner, Appl. Phys. Lett. 92, 033504 共2008兲. 14 D. K. Biegelsen, R. A. Street, C. C. Tsai, and J. C. Knights, Phys. Rev. B 20, 4839 共1979兲. 15 S. Dauwe, J. Schmidt, and R. Hezel, Proceedings of the 29th IEEE Photovoltaic Specialists Conference 共IEEE-PVSC-29兲, New Orleans, LA, 2002 共unpublished兲, pp. 1246–1249. 16 B. Hoex, F. J. J. Peeters, M. Creatore, M. A. Blauw, W. M. M. Kessels, and M. C. M. van de Sanden, J. Vac. Sci. Technol. A 24, 1823 共2006兲. 17 J. D. Kubicki and D. Sykes, Am. Mineral. 78, 253 共1993兲. 18 M. Park, C. W. Teng, V. Sakhrani, and M. B. McLaurin, J. Appl. Phys. 89, 1130 共2001兲. 19 A. Janotta, R. Janssen, M. Schmidt, T. Graf, M. Stutzmann, L. Görgens, A. Bergmaier, G. Dollinger, C. Hammerl, S. Schreiber, and B. Stritzker, Phys. Rev. B 69, 115206 共2004兲. 20 R. Janssen, A. Janotta, D. Dimova-Malinovska, and M. Stutzmann, Phys. Rev. B 60, 13561 共1999兲. 21 P. J. Rostan, U. Rau, and J. H. Werner, Proceedings of the 21st European Photovoltaic Solar Energy Conference and Exhibition 共EU-PVSEC-21兲, Dresden, Germany, 2006, p. 1181. 22 S. Olibet, E. Vallat-Sauvain, and C. Ballif, Proceedings of the 21st European Photovoltaic Solar Energy Conference and Exhibition 共EU-PVSEC21兲, Dresden, Germany, 2006, p. 1366. 23 E. Yablonovitch and T. Gmitter, Appl. Phys. Lett. 49, 587 共1986兲. 24 T. Mueller, S. Schwertheim, M. Scherff, U. Zastrow, Y. Huang, and W. R. Fahrner, Proceedings of the 22nd European Photovoltaic Solar Energy Conference and Exhibition 共EU-PVSEC-22兲, Milan, Italy, 2007, p. 1552.